Short answer: No general dictation tool ships knowing every drug name and ICD-10 code. The ones that work for doctors let you teach them. With a personal medical dictionary and replacement rules, you load your own medications, abbreviations, and code mappings so transcripts come out clean the first time.
Every clinician who has tried dictation has the same story. You speak a medication name and it lands as a garbled near-miss. You dictate a diagnosis and the engine has no idea you wanted the ICD-10 code that goes with it. After a few of these, you give up and go back to typing, which is slower and arguably the reason you wanted dictation in the first place.
The good news is that this problem is solvable, and the solution is not "find the one magic app that already knows medicine." It is to use a tool that learns your vocabulary. This guide explains why generic dictation fails on clinical terms, and exactly how to set up voice to text for doctors that knows your drug names and ICD-10 codes on a Mac.
Why generic dictation mangles drug names
Medication names are uniquely hostile to speech recognition for a few reasons.
First, many of them are invented words with no natural-language pattern to lean on. A general transcription engine is trained on ordinary speech, so when you say a brand name it has rarely encountered, it picks the closest common word that sounds similar. The result is a confident, plausible, wrong transcription.
Second, medicine is full of look-alike, sound-alike (LASA) drug pairs. These are medications whose names are close enough that even trained pronunciation can be ambiguous to a listener, let alone an engine. Patient-safety organizations maintain entire reference lists of these confusable pairs precisely because the risk is well documented. A dictation tool that guesses between two similar-sounding drugs is not a convenience, it is a hazard, which is why teaching the engine the exact set you prescribe matters so much.
Third, drug entries are rarely just a name. They come with a dose, a route, a frequency, and often a brand-versus-generic distinction. Speaking "twenty milligrams by mouth once daily" and getting clean numerals, units, and abbreviations out the other side is its own challenge for a general-purpose engine.
Why ICD-10 codes are even harder
If drug names are difficult, ICD-10 codes are a different category of problem entirely. A code like E11.9 is an alphanumeric string with a letter, digits, and a decimal point. Nobody speaks codes that way naturally, and no general speech engine is going to convert the spoken phrase "type 2 diabetes without complications" into the string E11.9 on its own. That mapping lives in your head and in the code set, not in the acoustic model.
So there are really two distinct needs here:
- Recognition: getting the engine to transcribe a medical word or a code string accurately when you say it.
- Expansion: letting you speak a short, natural trigger and have it expand into a longer, exact string, such as a full ICD-10 code, that you would never want to dictate digit by digit.
A dictation tool built for clinicians has to handle both, and it has to do it without you fighting the software.
The fix: a personal medical dictionary and replacement rules
This is exactly what Smart Vocabulary in Voice Keyboard Pro is for. It is a personal dictionary with replacement rules, which gives you two complementary controls:
- The dictionary teaches the transcription engine the words you actually use, so the medications, conditions, eponyms, and abbreviations in your practice get transcribed correctly instead of being replaced with the nearest everyday word.
- Replacement rules let you map a spoken trigger to an exact output string. Say a short phrase you choose, and out comes the full expansion, whether that is a properly formatted ICD-10 code, a standard order phrase, or a normalized drug spelling.
The key idea is that you are not hoping a vendor guessed your specialty correctly. You define the mappings you use every day, so the tool fits your practice rather than the other way around. The same approach helps any specialist with dense jargon, which is why we recommend it across our guides for doctors on Mac and medical residents learning to document quickly.
The dictation tools that work for clinicians are not the ones that claim to know medicine. They are the ones that let you teach them your medicine.
Setting up a clinical vocabulary, step by step
Here is a practical way to build a vocabulary that pays off within your first week.
1. Start with your top 50 medications
You do not need to load a pharmacopeia. Most clinicians prescribe from a relatively small working set. Pull the medications you write most often and add each one to your dictionary, including the brand and generic forms you use interchangeably. Add the spelling you want to see in the chart, so dictation produces the form your EHR and your colleagues expect.
2. Add your common abbreviations and order phrases
Set up replacement rules for the phrases you repeat dozens of times a day. A spoken trigger like "normal heart exam" can expand into your standard sentence about a regular rate and rhythm with no murmurs, rubs, or gallops. This is where dictation stops being a transcription tool and starts being a genuine time machine, because you are collapsing a long, repetitive sentence into two or three spoken words.
3. Map the ICD-10 codes you reach for
You almost certainly use a small number of codes far more than the rest. Create a replacement rule for each frequent one. For example, you might set a trigger so that a short spoken phrase you choose expands to the exact code:
- A trigger for type 2 diabetes without complications expands to
E11.9 - A trigger for essential hypertension expands to
I10 - A trigger for hyperlipidemia, unspecified expands to
E78.5 - A trigger for a routine adult exam with no abnormal findings expands to
Z00.00
You decide the trigger wording, you confirm the code against your own reference, and from then on the exact, correctly formatted string drops into the note whenever you say it. Build the list from the codes you genuinely use, and resist the urge to map hundreds you rarely touch, because a lean list is faster to recall and less error-prone.
4. Grow it as you go
Every time the transcript gets a term wrong, that is a signal, not an annoyance. Add the missed word to your dictionary in the moment and it stops being a problem. Within a couple of weeks, the corrections taper off and dictation starts feeling reliable for the language you actually use.
5. Decide how you want doses and units to look
Doses are where a lot of clinical dictation quietly goes wrong, because there is more than one correct-looking way to render the same instruction. Do you want "20 mg" or "20 milligrams"? "PO" or "by mouth"? "BID" or "twice daily"? The transcript should match whatever your chart, your specialty, and your colleagues expect to read. Decide once, then encode it with replacement rules so a spoken phrase like "twenty milligrams by mouth twice daily" produces the exact abbreviated form you prefer every time. Consistency here is not cosmetic. It reduces the chance that a hurried reader misreads an order, and it keeps your notes uniform across hundreds of encounters. Pick a convention, write the rules to enforce it, and stop making the same formatting micro-decision on every single line.
How this fits a Mac dictation workflow
Voice Keyboard Pro is a native macOS menu bar app. You hold a hotkey, speak, and release, and the text appears at your cursor in whatever app you are working in. For clinical use that matters, because EHR note fields, secure messaging, and your own documentation all become dictation targets without any copying and pasting. The cursor is already where the note needs to go, so the words land there directly.
Because the vocabulary travels with the app rather than living inside one specific program, your drug names and code mappings work the same whether you are in a browser-based EHR, a desktop notes app, or an email to a colleague. That portability is a real advantage over dictation that only functions inside its own window. If you want a broader look at the category, our overview of medical dictation software puts these options side by side.
Accuracy, review, and patient safety
A faster note is only better if it is also a correct note. Dictation, however good, does not relieve you of review, and with medications and codes the stakes are higher than with ordinary prose. Two habits keep you safe:
- Always proofread medication entries. Confirm the drug, the dose, the route, and the frequency before you sign. A loaded dictionary makes the transcription reliable, but you are still the clinician of record.
- Verify codes the first time you map them. Set each replacement rule against an authoritative code reference, double-check it once, and then trust the mapping. The error you want to avoid is a wrong code repeated automatically, so getting the mapping right at setup time is what protects you.
This is also why the privacy posture of your tool matters. Clinical dictation involves protected health information, and the requirements around it are specific. We cover that in detail in our companion guide to HIPAA-compliant voice to text for clinical notes, including why a signed Business Associate Agreement is the real foundation of compliant use. Read that before you dictate any identifiable patient information.
How much time does this actually save?
The math is straightforward. Most adults type around 40 words per minute and speak at 130 to 150 words per minute. For the narrative portions of a note, dictation is two to three times faster than typing once your vocabulary is dialed in. Layer replacement rules on top, where a two-word trigger expands into a full sentence or an exact code, and the effective speed for those repetitive elements is far higher still.
Across a full clinic day of twenty-plus encounters, shaving even a minute or two off each note adds up to meaningful time returned to you, often the difference between charting finished at the clinic and charting carried home. The point of teaching the engine your drug names and codes is not novelty. It is getting that time back without sacrificing accuracy.
Common questions
Does any dictation app know all drug names out of the box?
No tool reliably knows every brand, generic, and dose format for every clinician. The realistic and safer approach is a tool that lets you load the medications you actually prescribe, so the ones you use are transcribed correctly and the ones you never touch do not create false matches.
Can it insert ICD-10 codes automatically from a diagnosis?
It inserts the codes you have mapped. You create a replacement rule connecting a spoken trigger to a specific, verified code, and then it expands reliably whenever you say it. This is not automated coding, which is a clinical and billing judgment you make. It is a fast, accurate way to enter the codes you have chosen.
Will my vocabulary work inside my EHR?
Because the app types at the cursor system-wide, your dictionary and rules work in most EHR note fields, including the browser-based ones, as well as in any other Mac app. The vocabulary belongs to you, not to a single program.
What about specialties with very dense jargon?
The denser your terminology, the more a personal dictionary helps. Surgeons, oncologists, and psychiatrists all have niche vocabularies, and the same setup process applies: load the terms you use, add replacement rules for the phrases you repeat, and refine as you go.
The bottom line
Voice to text for doctors that knows your drug names and ICD-10 codes is absolutely achievable, but not by waiting for a tool that ships knowing your specialty. The clinicians who get reliable dictation are the ones who spend a focused hour teaching the tool their working vocabulary. With Smart Vocabulary's personal dictionary and replacement rules, that setup is quick, it lives system-wide on your Mac, and it pays back every single clinic day after.
Voice Keyboard Pro has a free tier, so you can load a handful of your most-used medications and codes and feel the difference on your next batch of notes. Build the vocabulary once, review your notes as always, and let your voice do the typing.